← ClaudeAtlas

gabe-reviewlisted

Code review with risk pricing, confidence scoring, interactive triage, and deferred item tracking. Also checks plan alignment (is this diff on-scope?), detects stale verified topics, and proposes DECISIONS.md entries for architectural changes. Surfaces the cost of NOT fixing each finding. Usage: /gabe-review [target] or /gabe-review deferred
khujta/gabe-suite · ★ 0 · Code & Development · score 63
Install: claude install-skill khujta/gabe-suite
# Gabe Review — Code Review with Risk Pricing ## Codex Command Bridge When Codex invokes this skill as the `Gabe Review` command surface, first read the active command wrapper from `.agents/commands/gabe-review.md`, `~/.agents/commands/gabe-review.md`, or `~/projects/gabe_lens/commands/gabe-review.md`, then preserve that command's argument routing and visible output contract. This `SKILL.md` is the review engine referenced by the command file; if there is any conflict, the command file controls command-time behavior such as `Gabe-Lens block` rendering, singleton `REVIEW.md` reconciliation, and mode-specific skips. ## Purpose Review code changes and price every finding — what it costs to fix now, what it costs to ignore, and what you're betting by deferring. Track deferred items across reviews and escalate when the same gap gets kicked down the road. This is NOT a generic checklist review. Every finding gets a **Defer Risk** (consequence + probability) and a **Maturity Gate** (MVP/Enterprise/Scale). The output is a risk matrix with a **Review Confidence Score** that lets humans make informed ship/defer decisions — and an interactive **Triage** loop to resolve findings on the spot. > **Rendering note.** Output templates in this spec wrapped in bare triple-backtick fences (no language tag) are spec-meta delimiters — render contents as plain markdown at runtime so findings tables display as tables, not monospace code. Tagged fences (```bash, ```diff, etc.) stay fenced at ru